Spatiotemporal Pattern of Neural Processing in the Human Auditory Cortex

See allHide authors and affiliations

Science  06 Sep 2002:
Vol. 297, Issue 5587, pp. 1706-1708
DOI: 10.1126/science.1074355


The principles that the auditory cortex uses to decipher a stream of acoustic information have remained elusive. Neural responses in the animal auditory cortex can be broadly classified into transient and sustained activity. We examined the existence of similar principles in the human brain. Sound-evoked, blood oxygen level–dependent signal response was decomposed temporally into independent transient and sustained constituents, which predominated in different portions—core and belt—of the auditory cortex. Converging with unit recordings, our data suggest that this spatiotemporal pattern in the auditory cortex may represent a fundamental principle of analyzing sound information.

One of the basic operations that the auditory system performs to decode acoustic information is the temporal analysis of sound features and their decomposition into specific neural discharge patterns. The importance of temporal analysis in audition is evident considering that the sensory stream emerges mainly in series over time. Decomposition and integration in the time domain are used for the qualitative and quantitative perceptual analysis of sound information (1,2), and neural mechanisms specifically dealing with temporal pattern analysis are essential for the processing of complex auditory signals. In animals, neurons of the auditory cortex can be broadly classified into transient and sustained responders. Transient responses typically occur at the onset of a stimulus, whereas sustained responses follow the stimulus (3–9). In the human auditory cortex, the spatiotemporal principles of encoding sound information are not yet fully understood. Magnetoencephalography demonstrated transient and steady-state responses (10), and neuroimaging showed a stimulation rate–dependent modulation of the temporal response pattern (11). Here we asked whether transient and sustained neural activity patterns in the human auditory cortex are elicited simultaneously by one and the same stimulus and whether there is a specific pattern of spatial distribution. We combined “magnetization-prepared” functional magnetic resonance imaging (fMRI) and acoustic stimulation by the gradient sounds to study the neural activation after a baseline of silence (9,12) (fig. S1). Further, we used independent component analysis (ICA) (13) in a hierarchical combination of “spatial” ICA (14) and “temporal” ICA (15) to blindly decompose the evoked blood oxygen level–dependent (BOLD) signal mixtures into constituent spatiotemporal sources (9, 16, 17).

Spatial ICA (14) extracted individually unique maps and associated time courses (Fig. 1) in the supposed primary and secondary auditory cortices (18). The time courses were characterized by an initial peak, then a plateau of persistent and irregularly oscillating signal effects superimposed. This interindividually consistent temporal pattern suggested the presence of at least two concurrent—possibly temporally independent—transient and sustained processes. However, the two presumptive processes concurred in the same spatial components and were mixed with other, less-consistent oscillatory phenomena. Assuming multiple, spatially overlapping, independent signal sources in each imaging voxel, we decomposed by temporal ICA (15) the signals from the spatial ICA regions into maximally temporally independent components (9) and identified a transient and a sustained component (Fig. 2). These concurrent components were blended with the signal mixtures arising from all fields of the auditory cortex. To identify portions with predominantly transient or sustained activity, we mapped the components to the cortical space (9). The distribution of response was characterized by a central stripe-like area (Heschl's gyri) predominated by sustained responses and a surrounding area (temporal and polar planes, temporal opercula) predominated by transient responses (Fig. 3).

Figure 1

Spatially independent components and associated time courses of the blood oxygen level dependent (BOLD) signal response to repetitive scanner sounds are projected on individual anatomical slices (radiological convention) positioned through the activation areas of individual subjects (S1 to S8). Spatial ICA blindly decomposed the presumptive primary and secondary auditory cortex. The associated time course was generally characterized by an initial peak at about 5 to 10 s after stimulation onset and evolved into a stationary plateau of activation. The initial transient phenomenon was highly consistent across subjects, whereas the sustained phase was associated with considerable interindividual variation and irregular oscillations. (Bottom right) The mean ± SE of the individual signals.

Figure 2

Temporal ICA of the signal behavior within the auditory cortex (Fig. 1). The temporally independent components of the BOLD signal time course are illustrated as image plots (top) (29, 30) and grand average (± SE) line plots (bottom) of all trials (n= 5) and subjects (n = 8). The transient and sustained temporal components showed considerable consistency across trials and subjects. Their most representative parameters (mean ± SD) were as follows. Transient: onset time (time to 10% of peak), 2.8 ± 0.19 s; time-to-peak (delay from stimulation start to maximum), 5.1 ± 0.16 s; peak width (time in which signal was ≥ 10% of peak), 4.7 ± 0.22 s. Sustained: onset time, 1.2 ± 0.11 s; rise time (time from 10% of peak to 90% of peak first crossing), 4.9 ± .89 s.

Figure 3

(Top) The relative contribution map of transient and sustained BOLD signal sources across all subjects and trials with the corresponding signals as identified by temporal ICA. (Bottom) Signals represent intraindividual averages of the five trials used as predictors within a group multiple regression analysis (31). The functional map is projected on the reconstructed cortical surface of the temporal lobes of a standard brain template. Color coding indicates the relative contribution of the two predictor classes and suggests a spatial continuum between the temporal response patterns. The contribution of the sustained response type becomes less predominant as one moves from the core to the belt areas. There was no notable hemispheric difference in the extension of the predominantly transient and sustained responses.

Human neuroimaging studies show transient or sustained temporal patterns of response to different types of acoustic stimuli (11, 19, 20). We demonstrate that the brain can use different temporal codes for one and the same stimulus. On the microscopic level (below the resolution of fMRI), the BOLD signal within an imaging voxel arises from tissue that is presumably composed of single, nearby sources of transient and sustained signals. We cannot determine whether the observed signals were produced by different or the same cell populations, whose response patterns follow a continuum from sharp phasic to robust sustained responses (4). Superposition of varying phasic events, possibly reflecting neural habituation and other adaptation processes, could be related to the irregular oscillations seen in the sustained phase. This finding is also consistent with the existence of two neuron populations in the monkey auditory cortex, one exhibiting stimulus-synchronized and the other nonsynchronized discharge patterns (8). On the macroscopic level, the predominance of transient and sustained signal sources was reminiscent of the parcellation into core and belt areas of the monkey auditory cortex (21,22), a distinction that appears to be paralleled in the human brain (23, 24). The specific functional operations associated with this spatiotemporal response pattern and possible hemispheric differences remain to be elucidated (25). Studies in monkeys and humans showed that the response preference to spectral complexity increases from the core to the belts (2, 23) and suggested that this might be related to “what” and “where” processing streams (26).

Because our imaging parameters range on different temporal and spatial scales than electrophysiological recordings (3–8,10), they could represent different levels of processing. Our data add to the evidence that the temporal decomposition of neural activity into transient and sustained patterns, or a continuum of them (4), may be a fundamental principle of deciphering auditory information. The mechanisms of upstream propagation of differential neural activity have only been partially unraveled. At the cortical level, the transformation into different temporal response types could be achieved by separate synaptic networks (27). The general rules, however, need to be viewed in light of the auditory network at large. In the thalamocortical circuitry, for instance, neural signals undergo radical reconstruction, with some properties preserving fidelity and others being transformed or generated anew in the auditory cortex (28). Temporal signal transformation appears to be a fundamental principle in the auditory system and could be related to different hierarchical levels of sound characterization. This hypothesis becomes particularly perspicuous considering the serial properties of auditory information.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1


  • * To whom correspondence should be addressed. E-mail: erich.seifritz{at}


Stay Connected to Science

Navigate This Article